P7 CPU Chip

  #1  
Old 01-14-2014, 04:46 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
P7 CPU Chip

While working for IBM, I developed, and taught a semiconductor (microelectronics) course. One of my "show and tells" was a P7 CPU chip (wafer, and modules). The P7 was developed by IBM in 2009, using a 45nm (nanometer) technology. 45nm is the minimum feature size (poly width) for the chip design. The P7 successor is the P8 CPU.

Here are some photos, that I thought would be of interest. The DOF is extremely shallow in some of the photos, even at f/22.


1 300mm silicon wafer. Red rectangle is one of many P7 CPU chips on the wafer.



2 Closer image of P7 CPU chip.



3 P7 CPU chip size comparison to penny.



4 Magnified edge of P7 CPU chip.



5 Magnified edge of P7 CPU chip.



6 Eye of needle used for size reference. Spherical shapes are solder ***** used for I/O, power, and ground connections to chip.



7 Black rectangle is backside of “diced” chip attached to ceramic substrate. Chip solder ***** are reflowed to attach chip to ceramic substrate.



8 Edge of ceramic substrate.



9 Bottom of ceramic substrate. Gold pads are for chip connections to socket/card.



10 Top of completed module.



11 Edge of completed module.



12 Bottom of completed module.
 
  #2  
Old 01-14-2014, 05:05 PM
JackandJanet's Avatar
Global Moderator &
Senior Member

Join Date: Aug 2007
Location: Among javelinas and scorpions in Zoniestan
Posts: 7,776
Received 51 Likes on 48 Posts
It's completely amazing to me that it all works.

Nice pictures, as always.

Edit: This may be a totally stupid/dumb question, but do you or anyone know why rectangular chips are produced on round wafers?

- Jack
 

Last edited by JackandJanet; 01-14-2014 at 05:09 PM.
  #3  
Old 01-14-2014, 05:17 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Originally Posted by JackandJanet
It's completely amazing to me that it all works.

Nice pictures, as always.

Edit: This may be a totally stupid/dumb question, but do you or anyone know why rectangular chips are produced on round wafers?

- Jack
Good question!

The wafers are sliced from a cylinder, which is round by the way it is fabricated. Chips are either square or rectangular, because their image is "stepped" across the wafer.
 
  #4  
Old 01-15-2014, 09:19 AM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Check out this link for P7 CPU specs: http://www.7-cpu.com/cpu/Power7.html
 
  #5  
Old 01-15-2014, 05:56 PM
JackandJanet's Avatar
Global Moderator &
Senior Member

Join Date: Aug 2007
Location: Among javelinas and scorpions in Zoniestan
Posts: 7,776
Received 51 Likes on 48 Posts
Originally Posted by Takeda
Good question!

The wafers are sliced from a cylinder, which is round by the way it is fabricated. Chips are either square or rectangular, because their image is "stepped" across the wafer.
I never thought to question the design of a chip. Using a rectangle or square shape maximizes surface area and, I suspect it works best with the shape of the internal circuitry.

And, I suppose a cylindrical wafer blank is easier to fabricate than a rectangular one. Still, I can't help but notice the wasted area around the perimeter, and it seems to me a rectangular wafer would be more efficient. But, I imagine the fab engineers have considered all this, rejecting my thoughts.

- Jack
 

Last edited by JackandJanet; 01-15-2014 at 06:00 PM.
  #6  
Old 01-15-2014, 06:13 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Originally Posted by JackandJanet
I never thought to question the design of a chip. Using a rectangle or square shape maximizes surface area and, I suspect it works best with the shape of the internal circuitry.

And, I suppose a cylindrical wafer blank is easier to fabricate than a rectangular one. Still, I can't help but notice the wasted area around the perimeter, and it seems to me a rectangular wafer would be more efficient. But, I imagine the fab engineers have considered all this, rejecting my thoughts.

- Jack
Jack, the silicon "cylinder" is pulled from molten silicon, while spinning. This is why it is round, instead of square. And a round wafer is easier to handle through the different processes.
 
  #7  
Old 01-15-2014, 06:21 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Jack, here ya go, I found a video of the wafer making process:

 
  #8  
Old 01-16-2014, 12:31 PM
JackandJanet's Avatar
Global Moderator &
Senior Member

Join Date: Aug 2007
Location: Among javelinas and scorpions in Zoniestan
Posts: 7,776
Received 51 Likes on 48 Posts
Thanks, Takeda. That clears things up very well. I suppose I could have found that myself if I had used my search finger, but like some others, I was just too lazy at the time.

It's still fascinating to me that a silicon chip can do so much, so reliably. I know that CPUs can fail, but they must be the last components to do so in normal use. Those tiny solder points are just amazing too.

I attempted to find information on the manufacturing failure rate of these components and didn't really succeed, possibly because the answer doesn't exist since chips may still be used if they are not "perfect" and actual yield rates may be confidential. However, I found this well-written article that's only 4.5 years old, so it should still be fairly relevant: http://www.geek.com/chips/from-sand-...s-made-832492/

Having taught digital processor/systems design for many years, I'm in awe of the physical design of modern ULSI circuits. What an amazing design/development process!

- Jack
 
  #9  
Old 01-16-2014, 11:25 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Originally Posted by JackandJanet
Thanks, Takeda. That clears things up very well. I suppose I could have found that myself if I had used my search finger, but like some others, I was just too lazy at the time.

It's still fascinating to me that a silicon chip can do so much, so reliably. I know that CPUs can fail, but they must be the last components to do so in normal use. Those tiny solder points are just amazing too.

I attempted to find information on the manufacturing failure rate of these components and didn't really succeed, possibly because the answer doesn't exist since chips may still be used if they are not "perfect" and actual yield rates may be confidential. However, I found this well-written article that's only 4.5 years old, so it should still be fairly relevant: http://www.geek.com/chips/from-sand-...s-made-832492/

Having taught digital processor/systems design for many years, I'm in awe of the physical design of modern ULSI circuits. What an amazing design/development process!

- Jack
You mention yield, and failure rate Jack. Two different things Jack. To yield is confidential, but typically, worst case failure rates are published. I'll see if I can find some information on this for you. Device infant mortality can be improved with burn-in processes, with a cost adder.
 
  #10  
Old 12-27-2014, 12:50 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
A new photo using extension tubes. Eye of the needle for size comparison:

 
  #11  
Old 12-27-2014, 01:22 PM
JackandJanet's Avatar
Global Moderator &
Senior Member

Join Date: Aug 2007
Location: Among javelinas and scorpions in Zoniestan
Posts: 7,776
Received 51 Likes on 48 Posts
Amazing, the uniformity of those tiny solder points!

- Jack
 
  #12  
Old 12-27-2014, 08:41 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Originally Posted by JackandJanet
Amazing, the uniformity of those tiny solder points!

- Jack
Thank you Jack.
 
  #13  
Old 12-28-2014, 12:07 PM
joe51's Avatar
Senior Member
Join Date: Oct 2011
Posts: 876
Received 14 Likes on 13 Posts
Originally Posted by JackandJanet
I can't help but notice the wasted area around the perimeter,
- Jack
Adding the insulating, semi-conducting and conducting layers and "features" is the real work when fabricating ICs so it's not as much of a waste as you might think since no features are added to those edge pieces. Also the raw silicon edge pieces can be melted and reused to create new wafers.
 
  #14  
Old 12-28-2014, 12:11 PM
Takeda's Avatar
Senior Member
Thread Starter
Join Date: Dec 2011
Location: Durham, NC
Posts: 443
Likes: 0
Received 0 Likes on 0 Posts
Originally Posted by joe51
Adding the insulating, semi-conducting and conducting layers and "features" is the real work when fabricating ICs so it's not as much of a waste as you might think since no features are added to those edge pieces. Also the raw silicon edge pieces can be melted and reused to create new wafers.
The newer semiconductor technologies uses S.O.I. (Silicon On Insulator) so the edge pieces cannot be melted down.
 
  #15  
Old 12-29-2014, 12:21 PM
joe51's Avatar
Senior Member
Join Date: Oct 2011
Posts: 876
Received 14 Likes on 13 Posts
Interesting. Have not kept up with some of the newer processes since retiring. How do you grow that on a crystal boule? Or is it created by another process?
 

Thread Tools
Search this Thread
Quick Reply: P7 CPU Chip



All times are GMT -4. The time now is 02:12 AM.